Beyond readability: Advancing patient-centered evaluation of AI-generated clinical reports.

Large language models offer promising opportunities to simplify clinical documentation and improve the accessibility of medical information for patients. Commenting on the recent article by Fereydooni et al. (2026) on GPT-4-generated patient-centered prostate cancer pathology reports, this correspondence argues that readability, while necessary, is not sufficient to ensure genuinely patient-centered communication. Drawing on recent evidence on after-visit summaries, discharge summaries, medical reports, open notes, and patient-facing educational materials, we suggest that AI-generated clinical documents should be evaluated through broader multidimensional criteria, including comprehensibility, relevance, usability, emotional impact, and empowerment potential. We also emphasize that patients and caregivers should be directly involved in the design and evaluation of AI-generated clinical communication tools. Moving beyond readability is essential to ensure that these tools do not merely simplify medical terminology, but support patients in making sense of health information and participating meaningfully in their care.